____twoism

Something I've noticed about a few of BoC's samples: They definitely had music in the background in their original forms, but in the BoC tracks they've somehow managed to trim it down to just the vocals. There's many examples of this, but a couple of the more baffling ones to me would have to be the lyrics from "Trapped" and the words from this Sesame Street bit (which have appeared in multiple songs):
[media]https://www.youtube.com/watch?v=QZyRR-18EVU[/media]

If I were to try and do the same for, say, the "Ohh!" at around 8 seconds into this clip, how would I go about doing it?:
[media]https://www.youtube.com/watch?v=AvW19WqbTeE[/media]

Trapped used the acapella version found on the 12". It clearly comes from an old vinyl source as you can hear the clicks and pops. Pretty sure this version is on youtube too. Same for Midas Touch as well.

With clever EQing you can eliminate background music whilst retaining the vocal.

Sounds can be also be isolated by splitting a stereo l/r track into 2 mono tracks and inverting the sound wave of one of them. Audacity can do this easily.

My limited understanding, based on a fair amount if homework a while ago, was that isolating vocals from a mixed track was actually very difficult and rarely that successful. Am I wrong? If yes, I'd love for someone to enlarge on that as there us a particular vocal I wanted to isolate for a project I was working on a while back and I wasn't entirely happy with the finished piece. Cheers.

Same goes for Many Me's, the video (that i uploaded btw haha) comes from a remastered tape with a lot of other edited vignettes, and Sesame Street added that jazzy tune (I read that somewhere).
Still, someone here spotted that almost every Sesame Street sample used by boc can be found in one episode, which contains the new version of Many Me's, if that's the case then they must have done something, maybe they re-recorded like they did with the speech in One Very Important Thought, but highly unlikely..

Peacock Tail wrote:Trapped used the acapella version found on the 12". It clearly comes from an old vinyl source as you can hear the clicks and pops. Pretty sure this version is on youtube too. Same for Midas Touch as well.

I was not aware that these existed.

WaitingFor_22 wrote:Same goes for Many Me's, the video (that i uploaded btw haha) comes from a remastered tape with a lot of other edited vignettes, and Sesame Street added that jazzy tune (I read that somewhere).
Still, someone here spotted that almost every Sesame Street sample used by boc can be found in one episode, which contains the new version of Many Me's, if that's the case then they must have done something, maybe they re-recorded like they did with the speech in One Very Important Thought, but highly unlikely..

I initially thought that ATP Sixtyten's intro sounded like a completely different kid, but upon closer inspection it's just been slowed down about 15%. Maybe a version without the music does indeed exist on the tape of episode 2852?

A full list of the episode content is available here 'http://muppet.wikia.com/wiki/Episode_2852'

https://youtu.be/ZHAtrnbu8ts

Orange. Although I suspect this has been found already.

Holy shit.
https://youtu.be/mpQXqnOStAs
I love you.

Both of those are quite well known

Yeah, sorry, we kinda beat you to the punch on that one.

Mexicola wrote:My limited understanding, based on a fair amount if homework a while ago, was that isolating vocals from a mixed track was actually very difficult and rarely that successful. Am I wrong? If yes, I'd love for someone to enlarge on that as there us a particular vocal I wanted to isolate for a project I was working on a while back and I wasn't entirely happy with the finished piece. Cheers.

I have probably a comparable amount of understanding, but I would agree this is what I'm aware of. It always baffles me how people do it.

And no I'm pretty sure EQing isn't a major factor in this process.

Mine matches both of yours. The only way I know of to isolate sound is to exploit a unique aspect of it vs others: frequency, amplitude, or channel presence. With most cases I've seen there's not enough uniqueness for clean isolation. And the picture usually changes along the duration too. Maybe there are engineering tricks out there, I'd just be surprised if there are any beyond exploiting as much uniqueness as possible and seeing where it happens to end up.

Skimming Stone wrote:Both of those are quite well known

And there I was feeling like I had just found a holy grail

Anyway. As far as vocal isolation goes it is close to impossible especially if you are using the finished product in a high quality studio mix. There is a certain ammount that one can do to drown out the vocals after partial isolation but this is clearly not the case imo.

Aerial Boundaries wrote:

Mexicola wrote:And no I'm pretty sure EQing isn't a major factor in this process.

I didn't say it was the major factor but it is an essential factor.

Here is the 'O' extracted from the Sesame Street video above at 08 seconds. It's the best I could do...

https://www.dropbox.com/s/bighys9bnalv6jo/O.wav?dl=0

Here's my attempt at extracting the Many Me's speech. As you'll hear, the most successful speech extraction are they 'bye, bye's' at the end. It could be why BOC used that part too.

https://www.dropbox.com/s/6swd1zl0ycud7bc/Many%20Me%27s.wav?dl=0

One Very Important Transmission https://www.dropbox.com/s/krzpaws19c1ocl7/OVIT.wav?dl=0

I'll stop now...

The way to do this is by using Audicity>Effect>Noise Reduction.

First, select a few seconds of the track that doesn't contain speech and press the 'get noise profile' button.
Then, highlight the entire track and go back to Audicity>Effect>Noise Reduction.
Select (for example) Noise reduction (db) to 48
Sensitivity to 14.00
Frequency Smoothing to 0.
Noise: Reduce
Then hit 'OK'.

Then repeat the process going through the entire track, selecting samples of the track where there is no speech.

I think these might sound a bit better with some more post-processing once the backing track is removed. I recall using some certain effects when I reconstructed Circle to make it sound a bit less harsh, but can't remember what they were.

(Here it is, in case you're wondering: https://www.dropbox.com/s/yqux6keaego1o ... d.ogg?dl=1 )

Seems like a good approach--automate as much as possible, then manually tune as needed. Audacity's noise filter is tricky though, it seems to go after lower frequencies more aggressively for some reason, the bottom end of voice can get lost quick.

only just realised that the many me video is used in whitewater

Unmixing a song is a lot like unbaking a cake. But if youre lucky you can sometimes do a reasonable job of scraping off the icing